Evidence to decision for PICO questions 4–7
Values and preferences
Shahaj, 2019(12): A range of individual and social factors including: familial (lack of support, need for separate meals), and environmental (sense of security, local amenities, healthy food availability) were identified as challenges to treatment adherence. Differences between clinicians’ and patients’ beliefs were potential sources of confusion and mistrust and were related to both cultural and individual beliefs (e.g. perceptions of symptoms, disease management, and treatment expectations).
Fragasso, 2012(6): Quality of life on antihypertensive therapy is an important issue because clinicians are asked to initiate drug therapy in mostly asymptomatic patients, who are never happy to become instead symptomatic, due to drug prescription.
Resources required
Luthy, 2008(78): Norvasc (amlodipine), which is a CCB), is one of the most commonly prescribed medications in the treatment of HTN. Even though amlodipine is now available as a generic, the cost is still significant. When taken every day as prescribed, patients without prescription benefits must pay, on average, USD 64.00 for a 30-day supply of the 10-mg dose (Drugs.com). The cost of amlodipine per pill equals about USD 2.13 (Drugstore.com, 2007). Coupled with the fact that patients with HTN commonly take many prescription drugs for other comorbidities, such as diabetes and hyperlipidemia, consideration of a patient’s economic situation while paying for many prescription drugs can be a major factor in determining patient compliance with the prescribed regimen.
Luthy, 2008(78): Thiazide: Hydrochlorothiazide (HCTZ), a thiazide diuretic, is a cost-efficient, first-line option when initiating treatment for HTN. For the commonly prescribed dose of 12.5–50 mg per day, patients can expect to pay about USD 12.00 for a 30-day supply, approximately USD 0.40 per pill.
Luthy, 2008(78): Captopril, an ACEi, is the most cost-efficient option for the adjunct treatment of HTN, costing USD 12.00 for a one-month supply of 50 mg tablets. The cost per pill equals about USD 0.22. While a captopril–HCTZ combination pill is available, the economic burden is similar to Norvasc at USD 2.00 per tablet. However, when administered separately, HCTZ and captopril are an effective and cost-efficient alternative to the popularly prescribed Norvasc.
Gu, 2015(8): Because medication costs are usually paid out-of-pocket by patients with HTN, local and national governments do not directly feel the impact of high drug costs. However, high drug costs likely have a big impact at the level of individual households and therefore indirectly on the national economy. Additionally, Chinese patients are reluctant to pay out of pocket for antihypertensive medications, and studies of Chinese patients have shown that out-of-pocket drug costs reduce medication adherence among patients with HTN and CVD.
Bramlage, 2009(79): Drug costs were highest for patients being treated and being persistent with ARB therapy (EUR 326.16), closely followed by patients persistent with CCB treatment (EUR 234.63) and patients switching between classes (up to EUR 268.07)
Chrysant, 2008(80): The primary endpoint was the cost of therapy, which declined by 33% and in this study resulted in a saving of USD 19.00 per patient/month after switching from a multiple-pill combination to a single-pill combination.
Cost effectiveness
Modelling studies were assessed for their overall quality by evaluating the structure of the model, appropriateness of the assumptions, sources of model inputs and sensitivity analyses. A formal quality assessment tool was not used. Most studies used a state transition model (Markov) model with variable cycle lengths (1 month to 1 year, variable time horizons (1 year to lifetime/95 years of age), almost all were from a payer/health system perspective. The majority did not use a systematic review to identify most appropriate model inputs; the method for choosing studies for utility inputs was seldom described. Most studies included one-way sensitivity analyses, some performed multi-way and probabilistic sensitivity analyses. An assumption that significantly influenced outcomes for Q 6–8 were that single-pill combinations increase compliance, resulting in lower BP.(14, 16) Costs due to adverse events were included by some,(20) not others. Although model and assumptions were extensively described in some studies,(8) inputs were not based on a systematic review, leading to the possibility of bias in choosing inputs. The generalizability of these studies is very limited due to contextual differences.
The cost effectiveness of low-cost essential antihypertensive medicines for HTN control in China was assessed in a modelling study by Gu and colleagues.(8) Based on a state transition model, cost effectiveness of treatment for stage 1 (BP 140–159/90–99 mmHg) and stage 2 (BP ≥160/100 mmHg) were calculated. One-way and probabilistic sensitivity analyses were used. Treating hypertensive adults with prior cardiovascular disease for secondary prevention was found to be cost saving in the main simulation and 100% of probabilistic simulation results. Treatment of all other patient cohorts, including sensitivity analyses, was found to be cost effective at a willingness to pay threshold of USD 50 000.
Data from 4500 US adults with HTN from the community quality index study were modelled to estimate cost and cost-effectiveness to payers of consistently providing the basic elements of BP management (visits and medications associated with recommended care)(81). They compared “usual care” with “improved care” (100% provision of recommended care processes). Some inputs were obtained from systematic review/meta-analysis of RCTs from the literature. Improved care cost USD 170 per person with mild HTN, USD 801 for moderate HTN and USD 850 for severe HTN annually.
Park and colleagues(82) conducted a systematic review of cost-effectiveness analyses of antihypertensive medicines. They included 76 studies in their review. These included 14 studies comparing medicines with no treatment, 16 studies comparing medicines with conventional treatment, 28 studies comparing medicines between medicine classes, 13 studies comparing medicines within medicine class and 11 studies comparing different combination therapies. Quality assessment was performed using the Quality of Health Economic Studies scale. Twenty-one studies scored >91, 10 studies scored <70 on a 100-point scale. 80% of reported funding was from industry, these studies provided positive evidence for the companies that sponsored them. The majority (41) of the studies were from Europe, 16 were from North America and 19 were from other countries. ARBs were the most frequently evaluated drug class (62 times either as intervention or comparator in 42 studies); the most frequently included ARB were losartan (20 studies) and irbesartan (15 studies). CCBs were the next most frequent class evaluated (32 times in 31 studies); amlodipine was the most common drug in this class (19 studies). Next in frequency were ACEi (28 studies) and beta-blockers (BBs) (25 times in 23 studies) (most common atenolol in 16 studies). Thiazide diuretics were evaluated 17 times, hydrochlorothiazide was most frequent (10 studies).
All antihypertensives were cost effective compared with no treatment (dominant USD 19 945/QALY). ARBs were more cost effective than CCBs in nine comparisons, whereas CCBs were more cost effective than ARBs in two comparisons. As previously noted, most of these were funded by industry and the results favoured the sponsor. ARBs were more cost effective than ACEis or BBs in all comparisons. Variations in study results are likely due to variations in study settings, analytic models, variations in cost and publication bias.
Using a state transition model, Tajeu and colleagues(83) study cost effectiveness of antihypertensive medications in white and black men and women in the United States. The simulation study population was modelled using demographic and clinical data from an ongoing observational study of risk factors associated with stroke (REGARDS study). Health states considered included stroke, coronary heart disease, heart failure, chronic kidney disease and end-stage kidney disease. The model included white and black adults with HTN and ≥45 years of age. Antihypertensive treatment was found to be cost effective with ICER/QALY of more than USD 10 000 for all groups.
In an economic analysis funded by Novartis of patients with chronic kidney disease treated with benazepril, the authors used data from the AIPRI study for transition probabilities in model inputs.(84) The rationale for selecting the other input sources was unclear. Benazepril was found to be dominant in the long run (7-year time horizon)
For low- and middle-income countries, Gad and colleagues conducted a cost-effectiveness analysis in the Ghanaian setting. Using a state transition model with six health states, one-year cycle length and a lifetime time horizon, they compared cost effectiveness of different classes of medications (ACEi, ARB, BB, CCB, thiazide-like diuretics, no intervention). All classes were found to be more effective than no intervention, thiazide diuretics were the most cost effective (GHS 276/DALY). CCBs were more effective and more expensive. ACEis, ARBs and BBs were less effective than thiazide diuretics. The results were maintained in sensitivity analyses.
Ekwunife and colleagues conducted a cost-utility analysis of antihypertensive medications in Nigeria.(85) They constructed a Markov model, with six health states, a cycle length of one year and a time horizon of 30 years, of patients stratified by cardiovascular risk. Probabilistic sensitivity analysis was conducted and results presented as cost-effectiveness acceptability frontiers. They found thiazide diuretics to be the most cost-effective option across cardiovascular risk groups, followed by CCBs, at a willingness to pay of at least USD 2000/QALY. The results were robust and insensitive to parameter alterations.
An industry-sponsored (Solvay Pharmaceuticals) cost utility analysis(86) comparing eprosartan with enalapril and nitrendipine found eprosartan to be cost effective at a willingness to pay threshold of EUR 30 000. Another industry-funded study (Pfizer)(87) from Taiwan comparing amlodipine and valsartan in a Markov model with a five-year time horizon and one-year cycle length did not include a systematic literature review for its inputs. This study found amlodipine to be dominant; the results were robust in sensitivity analyses.
A non-industry funded study in the Polish setting comparing ACEis and ARBs(88) found ACEis to provide improved outcomes compared to ARBs; the annual gain from change in treatment from ARB to ACEi for the Polish population was 830 QALY and 1018 life-years gained.
An economic evaluation sponsored by Daiichi-Sankyo in China(14) was very well designed. Model inputs for drug efficacy and other outcomes were based on a systematic review and MA/NMA. However, the process for selecting references for utilities was not clear, leaving the possibility of bias in choosing inputs for utilities. They constructed a Markov model with five health states, analysed from a payer perspective over a 20-year time horizon. Olmesartan/amlodipine single-pill combination was dominant, compared with Olmesartan and amlodipine multiple-pill combination and Valsartan/amlodipine single-pill combination.
Similarly, a single-pill combination of indapamide and amlodipine compared with multiple-pill combination therapy(16) was found to be cost saving in a Polish setting. The authors used a Markov model with eight health states, cycle length of one month, over a lifetime time horizon. These results were consistent in sensitivity analyses.
Lung et al compared cost effectiveness of a triple-pill strategy (consisting of amlodipine, telmisartan and chlorthalidone) with usual care based on data from a trial (TRIUMPH) incorporated into a discrete-time simulation model (10-year time horizon). They extrapolated the data for the proportion of individuals reaching BP target to disability adjusted life years (DALYs) averted. Modelling inputs were from the literature – no systematic review, no rational for the references chosen. Their results indicated a cost of USD 2842.79 per DALY averted over a 10-year period. Findings were robust to variations in all key-parameters.
Equity
Meiqari, 2019(11): Although beta-blockers (BBs), loop diuretics, and statins might be available in some community health systems (CHSs) in low-income countries, health insurance does not cover them at commune level. Patients seeking medication in the public sector face two problems. First, there is fragmentation and lack of consistency in prescribing medication between different levels. For example, doctors at higher levels may prescribe newer-generation medication that is not covered by health insurance at CHSs; if the patients want to keep using the same medication, they have to return to the higher-level facilities or purchase them at their own expense. While most basic HTN medication is cheap, newer generations may be less affordable. Second, current regulations, according to JAHR 2014,(89) allow provincial facilities to dispense HTN medication for short periods. These short periods of prescribed medication require more visits to health facilities, which increases the treatment cost for patients, reduces their compliance, and decreases the odds of HTN control.
Helmer, 2018(44): Much research has been done to assess the best hypertensive treatment approaches in black patients; however, there is a paucity of high-quality data. Although there are no published data assessing clinical outcomes specifically in black patients using ACEi or ARB monotherapy, evidence from subgroup analyses and cohort studies suggests that these patients may have higher rates of cardiovascular and cerebrovascular outcomes compared with those taking other antihypertensives
Tajeu, 2017(83): Increasing treatment rates and adherence among black adults may allow third-party payers and healthcare providers to align themselves with the National Academies of the Sciences – Health and Medicine Division’s commitment to address racial disparities in care.
Indirect evidence, Buckley, 2016(90): A unique and complex array of factors may influence African American beliefs about HTN. First, African Americans may have impaired access to health care and education, although one study suggested this gap has significantly decreased due to public health initiatives. Many African American participants expressed a distrust of the health care system and the belief that they received different or worse care than patients of other ethnicities. These beliefs may lead African Americans to choose alternative views on HTN. Alternatively, African Americans may have chosen to entrust friends, family, and community members with their medical care independently of their views on the medical system. Prior research has suggested that the immediate community significantly influences the beliefs and behaviours of African Americans.
Alsabbagh, 2014(91): Higher socioeconomic status (SES) was associated with a lower risk of nonadherence in 31 of 40 cohorts (77.5%), with no difference in one cohort, and with a higher risk of nonadherence in eight cohorts. Overall, the pooled adjusted risk estimate indicated a lower risk of nonadherence among individuals with a higher SES: 0.89 (95% CI 0.87–0.92; P 0.001). In health care research, low SES has proven to be a strong predictor of health care utilization, morbidity, and premature death. Nonadherence to chronic medications, such as antihypertensives (AHTs), can also be determined by low SES.
Lewis, 2012(92): Patients who experience fewer logistic barriers (i.e. difficulty obtaining clinic appointments and health insurance) have better medication adherence rates.
Acceptability
Shahaj, 2019(12): Deliberately choosing to avoid or reduce medication (intentional nonadherence), rather than forgetfulness, was a theme in some studies. For some patients, symptoms acted as a guide for the seriousness of their HTN and guided their medication use; for example, they stopped treatment if symptoms disappeared. Some were guided by stress, using medication to manage worry or anxiety rather than HTN. Fear of dependency affected the amount of medication they took.
Gwadry, 2013(93): A significant improvement in medication adherence was found with increasing age and provider visits, and reductions in multiple-dosing regimens and medication class.
Alghurair, 2012(94): Twelve surveys studied poor adherence caused by therapy-related barriers. The most commonly identified barriers from this dimension were occurrence of side-effects, complexity of drug regimens, and interference of medication taking with daily routines.
Wetzel, 2004(95): An inverse association between dose regimen and compliance is shown, with mean compliance percentages being higher on a once-daily regimen (85–94%) compared to a twice-daily regimen (75–88%).
Feasibility
Angeli, 2012(96): The use of single-pill combinations implies less flexibility in modifying the doses of individual components and the exposure of patients to unnecessary therapy. Moreover, should a patient develop side-effects to one component, the entire combination should be discontinued and replaced by multiple pills. Using single-pill combinations, the physician cannot easily titrate one component without changing the other. None of the tablets currently available on the market are able to be broken to allow sufficient flexibility. Only specific manufacturing options might be suitable to achieve a successful titration in clinical practice.
Outcome utilities
Please refer to below.
Table 78Utilities per outcome for PICO questions 4–7
View in own window
| Outcomes | Utility | Systematic review (SR) | Primary studies reported in the SR |
|---|
|
Hypertension
| 0.96 | Ren 2020(14) | Li 2015(15) |
| 0.98 (range: 1–0.95) | Kawalec 2015(16) | Burstrom 2001(17), Sullivan 2008(18), Wang 2008(19) |
|
Type 2 diabetes mellitus
| 0.985 | Gad 2020(20) | Salomon 2012(21) |
|
MACE
| Time NR: All CVD excluding stroke: 0.73 (95%CI: 0.69–0.76) | Kawalec 2015(16) | Lunde, 2013(22) |
|
Stroke
| First month after onset: 0.55 | Ren 2020(14) | Li 2015(15) |
| Days 1–3: 0.70 | Gu 2015(8) | Salomon 2012(21) |
| Days 4–28: 0.88 | Gu 2015(8) | Salomon 2012(21) |
| Chronic state: 0.65 | Ren 2020(14) | Huang 2017(23) |
| Time NR: 0.70 (95%CI: 0.67-0.73) | Kawalec 2015(16) | Golicki 2010(24) |
|
MI
| First month after onset: 0.60 | Ren 2020(14) | Li 2015(15) |
| Days 1–3: 0.58 | Gu 2015(8) | Salomon 2012(21) |
| Days 4–28: 0.94 | Gu 2015(8) | Salomon 2012(21) |
| Chronic state: 0.70 | Ren 2020(14) | Huang 2017(23) |
| Time NR: Disability weight 0.124 | Gad 2020(20) | Salomon 2012(21) |
|
ESRD
| ESRD pre-dialysis: 0.73 (95% CI: 0.62–1) | Cooper 2020(25) | Jesky 2016(26) |
| Hemodialysis: 0.75 (SD: 0.25) | Cooper 2020(25) | Briggs 2016(27) |
|
Cognitive impairment/dementia
| Patient rating: 0.85 (SD: 0.19) | NA | Rowen 2015(28) |
|
Patient rating:
mild dementia 0.79 (SD: 0.22)
moderate dementia: 0.72 (0.23)
| NA | Orgeta 2015(29) |
|
Carer rating:
mild dementia 0.63 (SD: 0.27)
moderate dementia: 0.52 (0.27)
| NA | Orgeta 2015(29) |
|
HF events
| First month after onset: 0.63 | Ren 2020(14) | Li 2015(15) |
| Chronic state: 0.73 | Ren 2020(14) | Huang 2017(23) |
| Time NR: 0.79 | Gad 2020(20) | Salomon 2012(21) |
|
Adverse events
| Common: 0.88 | Gu 2015(8) | Clinical Judgement |
| Infrequent: 0.70 | Gu 2015(8) | Salomon 2012(21) |
